Won’t be fooled again? Ezekiel Di Paolo and Enactive Autopoiesis: Semiology of Autopoiesis (VII)

I have distinguished between strict and loose definitions of the concept of autopoiesis. These definitions can overlap, in the sense that one might share conditions with another. They can also be inconsistent, internally and with each other, when they involve contradictory conditions.

A looser definition has more conditions, but these can be taken flexibly and vaguely. It might be odd to label something strict, when it involves fewer conditions, but the important point is that they must be applied strictly and consistently, in contrast to looser versions.

Loose definitions are liable to generate more inconsistencies, applying to varied and vaguely defined actual cases. Vagueness and variation are key here. Strict definitions can have very wide scope, but that scope imposes tight restrictions on actual cases, whereas loose definitions can be more inclusive, but at the cost of unclear, ill-defined or open-ended correspondence between concept and actual cases.

Broadly, a loose definition approach to autopoiesis involves a mapping of empirical cases, accompanied by general conditions attempting to justify them as autopoietic. A strict approach is rational and logical, outlining a consistent and rigorous version. This may or may not correspond to all or to any empirical cases, but it serves as a rational test for them.

I raise these distinctions for two reasons: first, to consider a general objection to my position, coming from Ezekiel Di Paolo’s work on enactive autopoiesis; second, to follow the implications of this critical discussion for a semiology of autopoiesis.

The objection denies the distinction between strict and loose, replacing it instead with the idea of bare and expanded definitions of autopoiesis: ‘… it is possible to expand the conceptual reach of the theory of autopoiesis by introducing the idea of adaptivity.’ (Di Paolo, ‘Overcoming Autopoiesis’, p 51)

The aim of this theoretical expansion is explanatory, not in the sense of explaining the theory, but rather of increasing its explanatory scope: how much it encompasses under a connected set of reasons.

The expansion is described through the metaphors of ‘softening’, ‘holding dear’, ‘fertility’, ‘radicality’ and ‘richness’. Construction is understood as ‘elaboration’ and ‘complementation’:

We must be able to criticize autopoiesis constructively and soften up its edges so as to create a more fertile theory that holds dear to the most radical and richest ideas of the original formulation and at the same time allows for the fact that these ideas do not explain everything and must be elaborated and/or complemented in order to be useful for the study of cognition (and social systems).

Ezekiel A. Di Paolo (2009) ‘Overcoming Autopoiesis: An Enactive Detour on the Way from Life to Society’ Autopoiesis in Organization Theory and Practice Advanced Series in Management, 43–68, p 46

Di Paolo further expands the bare definition from Maturana and Varela with concepts of autonomy, sense-making, embodiment, emergence, experience and precariousness. These lead to the idea of an enactive autopoiesis: autopoietic processes bringing about the worlds they are in – and themselves – through dynamic interaction.

The idea of conceptual fluidity justifies Di Paolo’s expansion. If the original and bare concepts are fluid, then it is possible to modify and enlarge them to respond to problems and objections.

As these ideas are applied to increasingly diverse domains, from psychology and psychiatry to ethics and the arts, the enactive approach itself continues to evolve, and its core concepts change and broaden as they are confronted with new problem areas that they were not originally intended to address. This leads to a certain fluidity in the core concepts that raises both scientific and pedagogical challenges for the further development of enaction.

Randall D. Beer, Ezekiel A. Di Paolo ‘The theoretical foundations of enaction: Precariousness’ BioSystems 223 (2023) 104823, pp 1-10, p 1

The metaphors used by Di Paolo to describe his method as fluid constructionism reveal an allegiance to a founding idea, ‘holding dear’ to original autopoiesis, and a desire for the theory to be effective, for it to bear fruit and be a rich explanatory framework with greater reach.

In order to achieve these twin aims, the theory must grow from its bare concepts. This highlights the importance of the organic, political and moral metaphor of ‘radical’ for Di Paolo’s method. It combines root and bare foundation with an uncompromising commitment to expansion and innovation.

Fluid conceptual construction grows through complementarity, or the addition of new parts; through elaboration, or the detailed development of current parts, whether minor or major, overlooked or central; and through expansion, or the demonstration of its applicability to new areas.

There are risks to such expansions. When a system reaches its upper limits in terms of sophistication, all three approaches – complementarity, elaboration and expansion – put stress on current designs and models.

Di Paolo defines precariousness as a productive and yet fragile factor of creative development. However, in many designs the creative development of an original model leads to a different kind of precariousness, a steady and sometimes catastrophic increase towards failure.

With expansion it becomes harder to ‘bolt on’ a different part without interfering with delicate interactions. Current parts reach their design maxima; like too many piston engines on a single craft. There are diminishing returns in finding as yet unexploited fields.

Philosophies also become cumbersome over time, weighed down by specialist tinkering and perverse professional effects, such as unyielding existential commitments, inertia, ambitions, financial incentives, and public and political expectations.

The flawed and balefully named ‘Peacemaker’ bomber took piston engine size and number to their limits (336 spark plugs had to be changed after each flight, for instance)

The precariousness or fragility of a theory echoes Di Paolo’s reference to both terms in his definition of autopoietic processes. Though he doesn’t make this link, his theory can be seen as an instantiation of its own descriptions. The theoretical evolution he advocates is precarious and thrives through a fluid conceptual growth through innovation.

The analogy between theoretical evolution and fragile autopoietic evolution is instructive for the extension of autopoiesis to minds, persons and social systems. It draws attention to the extensive networks that any theory, group or single practitioner depends upon. Analogies of this kind only provide clues, rather than any kind of proof, but the complexity of extended networks is a potential problem for autopoiesis viewed as social.

Even the most insular theoretician is far from autonomous, given dependence on antecedents, common languages and data, shared techniques and technologies, laboratories and field studies, observed subjects and populations, community validation, junior and senior collaborators, teachers and pupils, lines of communication and critical interactions.

Some of the most egregious cases of scientific bias and malpractice come from the denial of these dependencies and from misleading claims to autonomy, often at the expense of the work of others. Influenced by convoluted external relations, the desires, ambitions, dreads, self-regard and self-loathing that drive selfish acts do not accord well with ideas of autonomy.

For bare or strict definitions of autonomy, the solution to networks of dependencies is to consider any network in terms of ‘couplings’, where autopoietic systems are coupled together yet preserve their autonomy because their interdependence is regulated to keep it to specific interactions.

Regulation can apply to natural laws or to social agreements. It can describe an independent subset of laws and rules, or a necessarily interconnected frame, without exceptions or protected areas. This latter holism or universal interconnection is a threat to the idea of coupling, since any couple will be an abstraction from the whole and not a legitimately independent operation.

Coupling is adopted by Di Paolo, but in a refined form as evolving over time and as regulated through agency and behaviour. In many cases and theoretical uses, coupling doesn’t have to limit itself to couples; it can be multiple. As my argument progresses, I will contrast interdependent but autopoietic couplings with deeper power relations underlying such specific and well-ordered exchanges.

For these relations, autopoietic processes are surface appearances and autopoietic theory is constructed through abstraction from them, like a rational consciousness ignoring its unconscious social and psychological settings.

Against the autonomy required for autopoiesis, the concept of power is a better explanatory candidate for such relations, with their tensions and conflicts and their often concealed, disorderly and distant effects.

A rejoinder to my move towards underlying relations and power could rely on how fragile and complex systems can evolve from simple and rational rules. Beer and Di Paolo demonstrate this in a study of precariousness and Conway’s Game of Life:

Although the general possibility of gliders in GoL are of course implicit in the Conway physics, the existence of a particular glider here and now is not given a priori; it requires the right local conditions for the closed network of process dependencies to arise and be maintained. Rather than being reprogrammed, gliders are systemically precarious because they can emerge from (and decay back into) an underlying artificial chemistry.

‘The theoretical foundations of enaction: Precariousness’ p 4

Gliders are used as a ‘toy model’ to test and exemplify the theory of precarious autopoiesis, but Beer and Di Paolo acknowledge such models only provide ‘concrete illustrations’. The problem is that they do so by abstracting from different empirical cases and by replicating a theory in a simplified form. Such models cannot validate explanatory superiority.

This counter argument leads to another rejoinder. It denies that the theory of autopoietic precariousness is abstract. On the contrary, since the theory turns on enaction, it is the opposite of abstraction and autopoiesis is an evolution in dynamic relation to its environment.

In the following points Di Paolo describes how enaction adds to theories of the extended mind (EM) by avoiding some of the theoretical abstractions that make them inaccurate models:

We can list some of the contributions of the enactive
approach to the EM story.

  1. A workable, unbiased definition of cognition.
  2. A distinction between relational and constitutional domains.
  3. The notion of autonomy as a self-sustaining identity under precarious conditions.
  4. The grounding of sense-making and normativity in the process of adaptive identity generation.
  5. The distinction between coupling and asymmetric regulation of coupling leading to a definition of agency and behaviour.
  6. A workable concept of mediation (as the marker to follow in locating complex forms of identity generation).
  7. The notion of a topology of regional identities.
  8. A hint towards the role of social interaction and participatory sense-making in estabilizing the socio- linguistic cognitive self.
Ezekiel Di Paolo ‘Extended Life’ in Topoi (2009) 28:9–21, p 20
DOI 10.1007/s11245-008-9042-3

As responses to an accusation of abstraction, the points insist on an absence of theoretical bias prior to the empirical nature of mind; in particular, on avoiding separation of mind and environment. An enactive approach knits mind and environment together in constitutive action and behaviour. This is a two-way process and domains are themselves constitutive of actions and behaviours.

Since precariousness in an environment is a condition for identity and autonomy, no enactive system is abstracted from its environment. This is also true for acts of sense-making and for the creation of norms. Enactive autopoiesis therefore leads to a more subtle definition of coupling. It is linked to agency and behaviour. It depends upon mediation, local topologies and social interactions.

Even if enaction avoids stark kinds of abstraction in the way it connects evolution to circumstances, that’s not the problem. The abstraction doesn’t stem from the intuitively attractive way enactive processes connect environmentally. It comes from the insistence that these connections can be understood as self-sustaining and autonomously precarious.

For Di Paolo, autopoietic systems must be inherently prone to fail as they evolve within an environment. Paradoxically, this weakness is their strength. Autopoietic systems aren’t fragile because they can be destroyed from the outside, but rather because how they create and maintain themselves is inherently unstable and prone to failure.

Autopoiesis must be precarious for two connected reasons. First, the empirical cases Di Paolo wants to encompass are precarious as they evolve. Observed biological organisms, minds and social organisations survive and evolve in fragile ways, where precariousness indicates degradation and evolution over time, rather than an instantaneous living-then-destroyed passage.

Second, precariousness is an essential feature of autopoiesis defined as an unstable evolution. Instability is the result of self-creation, or having to risk new solutions to different internally and externally generated problems. It is the result of articulating individually and collectively unstable processes, such as an evolving coupling that calls for the invention of new rules, habits and behaviours.

Precariousness is not only an extension of more strict versions of autopoiesis, making it possible for them to explain a greater number of empirical cases. It also solves problems for the more strict definitions with respect to time and destruction.

If destruction corresponds to two completely separate states, it makes no sense as a breakdown over time. To be a timed process, destruction must draw a stretch of time together, like dread foreshadowing death, or steady wear building up to abrupt failure. Without such connections, there is no destruction, only a passage from one state to another unrelated one.

In an approach with interesting and surprising connections to existentialism, given the formalism of bare or strict autopoiesis when compared to existentialist phenomenology, Di Paolo connects and intertwines function and destruction, making them relative to one another along a precarious evolution, as in Sartrean angst or Heidegger’s being-towards-death.

This combination of autopoiesis, precariousness and time reconnects autopoietic theory to the experience of time in an existential sense. The phenomenological link is reinforced by Di Paolo’s debt to and interest in Hans Jonas, Maurice Merleau-Ponty and Michel Henry.

What has all this got to do with my opening image, with Charlie Brown’s misadventures at the hands of Lucy, as she pulls away the football for the umpteenth time? What has his doomed determination not to be fooled again got to do with enactive autopoiesis?

Part of Charles Shulz’s cartooning humour stems from an insight into precarious introspection, Charlie Brown’s existence is defined by a fragile yet also headstrong search for meaning and success against setbacks and foes. The opposite to Chaplin’s savvy tramp, he likewise doesn’t give up as he overcomes inner doubts, but only to be fooled again, rather than learn to deceive others.

This introspection and its relation to his other bodily strengths and weaknesses are Charlie’s identity. But he isn’t autonomous, nor is any comic hero caught in social pitfalls and traps. Lucy is manipulating his naivety and failings. That’s the other side of slapstick humour: it is also knowledge of human frailties allied to cruelty. For Charlie Brown and the Tramp, others are their weakness and their defining fortitude.


Even if if we extend the autonomous autopoietic system to include couplings (the Tramp and the Kid; Charlie and Lucy) the problem of lack of autonomy returns around the limits of any given coupling. The Charlie-Lucy coupling isn’t autonomous. It depends on a wide selection of characters, each one influencing how all the others interact and develop. Once processes belong to networks of dynamic interactions, no part of those networks can be autonomous, without severing it from further relations it depends upon.

Strict and base definitions of autopoiesis have two resources to solve the problem of limits. First, the definition of autopoiesis can be in terms of unity. An individual or a coupling will have a defining unity according to which which all its autopoietic processes operate. Second, autopoiesis can be defined in terms of a protective boundary or membrane containing all necessary processes.

Neither of these options are good candidates for Di Paolo’s version because of the requirement for evolution, where both unity and boundaries change over time. He rarely refers to boundaries, preferring the concept of operational closure and does not include unity in his list of essential characteristics of autopoiesis, though it returns as a consequence of networks.

Unity is replaced by the idea of precarious autonomy inconsistent with unchanging unity over time. Precariousness implies imbalance and passages of indeterminacy, whereas unity calls for well-determined relations and components. Precariousness takes precedence over full unity of a system and unity becomes a property of a core network overseeing changing components.

If precarious enaction is a dynamic interaction, where external characters, actors or forces have control over internal evolution, then enaction and autopoiesis are contradictory, because no externally controlled process can be autonomous. If enactive processes are fragile, as controlled by independent external actors, then precariousness and autopoiesis are incompatible.

No externally manipulated precarious process can be autonomous (even you, Snoopy).

This critical point can be broadened. Precariousness is a property of extended relations. It is a property of the whole, not of independent parts. It doesn’t come from within but rather explains how and why an inside is shaped in unpredictable ways by an outside.

A social system such as a company might have a series of protocols and mission statements, staff rules, manufacturing and financial capital, plans and projects. These could be claimed to define the firm’s autonomy, but its evolution will have been generated by external competition, chance events, changes in markets, fluctuations in supply and demand, alterations in staffing, and technological developments.

Business leaders might say they won’t be fooled again, but historical evidence shows a constant churn of strategies and approaches driven by unexpected events. If protocols, plans and internal communication are to be part of the autopoietic processes of the company, then the fact that they are regularly torn up and reviewed due to external circumstances implies that conditions for autonomy cannot be satisfied.

External pressures dictate the evolution of firms. This isn’t a question of autonomous choice, but rather of being forced to change in a certain way; for instance, by a successful competitor, a different legal status, or economic events such as interest or exchange rate fluctuations. Even strategies designed to insure against such events have turned out to become unforeseen pressures; for example, in crises driven by derivatives.

For a system to not be fooled again, there has to be a process of adaptation rendering it immune to an earlier threat. However, if adaptation is by population, then any given system is not autonomous as a unity but as a member of that population.

Even if we expand unity to an autopoietic population, there is still no autonomy, if an interaction with an environment explains adaptation, such that any changes can’t be justified according to autonomous actions and changes by the population, but rather through environmental influences.

In the case of social systems such as firms it is not only the case that transformative external events deny autonomy. It is also that solutions to the problems brought about by those events come from external sources. There are no clear limits to those sources either. Emulation, negation, absorption, inspiration, learning and creative serendipity take an apparent autonomy beyond itself and connect it to its environment for survival and evolution.

These objections are made stronger by the appeal to precariousness. If we grant that evolving autopoietic systems must be precarious, because they are innovating at system and component levels in response to internal and external threats, thereby making them necessarily unstable, then that instability and innovation run counter to unity and autonomy. It would make more sense to describe an unstable and riskily changing system as breaking down in unity and lacking the self-command required for autonomy.

Di Paolo’s account of enactive autopoiesis seeks to refute the above arguments. It has an explanatory facet and a justificatory one. Sense-making, embodiment, emergence and experience explain how an enactive process evolves. Autonomy and precariousness characterise those processes, but also justify how they are autopoietic.

Instead of falling back on to the most strict versions of autopoiesis, ones that are autonomous to the point of independence from an environment and ones where the processes are as stable and immutable as possible, he adopts more flexible concepts of autonomy and self-creation, such that any threat is not so much repelled as included in the concepts and the functioning of autopoietic systems.

As I have indicated in my example of Charlie Brown and Lucy, and in earlier posts about autopoiesis, the problem with these looser versions of autopoiesis is their dependence on ideas of autonomy and process that are vulnerable to external control. If a process is manipulable from the outside, it is more accurate to define and study it as allopoetic – created partly by external sources – rather than autopoietic.

It could be objected that potential external control is an invalid objection, because Di Paolo’s theory has nothing to do with willed control and because potential control does not demonstrate actual failure of autopoiesis, but merely a possibility under certain circumstances, where there is always the option of stating that in such cases autopoiesis merely stops or breaks down.

Against these counterpoints, I argue that control is a special case and a way of illustrating a wider argument. Control is a local and narrow manifestation of something much broader: power, or the ability to influence events, things and beings. In addition to systems, any process can also be analysed in terms of power relations and structures of power. These are complex because they touch on all levels and types of process, since the ability to influence takes many forms, works in many ways and at very different degrees.

Lucy is adept at manipulating Charlie’s hopes, insecurities, body, skills, social relations, desires, language, place in social practices such as sport, family and schooling. Power is a description of the relations supporting such manipulation. It explains a multi-directional set of different perspectives, drives and aims in opposition to one another, but also capable of productive alliances.

With its emphasis on autonomy and selfhood, autopoiesis takes an inner-outer stance: a precarious autopoiesis in a changing environment. Power describes a changing lattice of relations: a network of changing zones of influence with varied outcomes on constituent regions, themselves selected for different and contingent reasons when considered from other viewpoints.

There are retorts to my turn to power. First, power fails the law of parsimony. It is not needed to explain autopoietic processes. Power doesn’t do anything; it is a vague description of forces, not the physical forces themselves. Second, instead of that general idea of power, theories of autopoiesis have the scientific concept of energy, as described in thermodynamics.

Worries about energy sources have been a constant in autopoietic theory. Beer and Di Paolo discuss the recurrent problem of the need for external energy for autopoietic processes in terms of ideas of fragility and thermodynamics:

Another more restrictive kind of precariousness can arise when operational closures are physically instantiated: thermodynamic fragility. Thermodynamically fragile operational closures are those whose very form is dependent on thermodynamic conditions. For instance, if the organization requires strict temporal relations between processes that depend on reaction rates, or energy flows, etc.

Randall D. Beer, Ezekiel A. Di Paolo ‘The theoretical foundations of enaction: Precariousness’ BioSystems 223 (2023) 104823, pp 1-10, p 6

The appeal to thermodynamics rather than to a wider concept of power is justified for biological autopoiesis; that is, when an identity is assumed with external energy requirements that can be described and tracked. Beer and Di Paolo consider biological examples thermodynamically, where ‘the very organization of living systems is shaped by their need to resist the second law.’

Systems are precarious because their processes are dependent on the availability of energy as resistance to entropy. Fluctuations in energy render internal processes fragile yet also define them; for instance, when changes in the availability of energy determine dormant states like hibernation or seed dormancy. Though fragile, they remain autopoietic because precariousness is internal and they continue to regulate themselves as fluctuating systems.

Self-regulation is the weak point of this argument as an explanatory model. There is greater explanatory potential in observing regulation across a network that includes so-called autopoietic systems, because precariousness and regulation are better understood as features of the network as a whole and at many different and changing levels. Hibernation and dormancy are properties not only of individuals but of whole ecosystems; in fire-released seed dormancy and ecosystem regeneration, for instance.

Due to autonomy, autopoietic explanations must be local. Even if they take account of an entire environment (as a changing data set, for instance) this information will be taken from an internal perspective of operating limits, sustainability and ability to evolve. Holistic explanations deny autonomy, replacing it with interconnections that cannot be severed or concentrated on one part of a system without explanatory loss.

Taken holistically, precariousness comes from fluctuations in power across a network, not from the internal requirements of a portion of the network. This can be demonstrated through examples of system-wide antagonism. A multi-agent conflict is misunderstood if it is broken down into binary relations between autonomous enactive actors and their environments, because each binary relation is also in play in all the others.

For the effect of climate change on hibernation and migration, where there is a case of fragility brought about by dependence on fluctuations in energy, we can consider an individual animal or a species as autopoietic, setting up a study of the effects of warmer winters in terms of a binary relation between energy changes and effects on autopoiesis.

This will be a limited and skewed analysis of the reasons for those effects, since there will be wider influences missed when the binary relation is abstracted from the whole network; for example, where anthropogenic influences ‘modify fire-prone ecosystems.’ (Juli G. Pausas and Byron B. Lamont, ‘Fire-released seed dormancy – a global synthesis’ Biol. Rev.(2022),97, pp. 1612–1639.1612doi: 10.1111/brv.12855. p 1632)

A counter to this holistic approach could be to claim that system wide explanations and reasons are not determining for autopoiesis. There are system wide regulations and laws, but while obeying them an autopoietic system has its own determining regulations and precariousness and these define it as individual.

For autopoiesis, the significant fact is that an environmental datum has changed and this change influences autopoietic processes. Why that change has occurred is immaterial. If a seed is not woken from dormancy by fire, it’s irrelevant to the seed’s processes that the lack of fire was due to human cultivations and their reasons.

This answer doesn’t work for autopoietic processes that can analyse and react to reasons, or that are influenced by them yet do not have full grasp or control of them. If reasoning belongs to internal processes but introduces external influences to their evolution and regulation, then the internal processes aren’t autonomous and the system is not autopoietic. For these cases, power and holistic explanations are more accurate and useful, even from an internal perspective.

In terms of explanation, there is a helpful analogy with currencies. It is possible to study movements between two currencies; for instance, the Sterling-Euro coupling. However, this limited study misses the influence of third parties on both those currencies. An explanation failing to take account of the full network of currencies is liable to miss causes and influences. It is at risk of misdiagnosing movements; for example, by attributing a change to the Sterling rate to weakness against the Euro zone, when the root cause is Dollar movement against Euro dragging Sterling up or down.

For all cases, binary precariousness can be opposed to systemic precariousness. The first can be understood, for instance, as energy requirement: this autopoiesis requires energy in these amounts and fluctuating in this way. The second can be observed in power relations: all these systems, taken as a whole, are interdependent according to these power relations.

The reliance on binary precariousness is more serious when Di Paolo and Beer extend autopoiesis to mind and social systems. For such systems, a strict thermodynamic analysis is a blunt explanatory account when compared to power; in particular, if the explanations include assumptions about the autonomy and precariousness of minds and social systems in complex antagonistic situations.

The following map indicates world oil reserves per country in 2017. We could study social and political problems and developments around the map in terms of thermodynamic requirements and supply at country, business, production, population, class, household and individual levels.

To understand how and why relations have evolved, energy supplies and flows have much less explanatory effectiveness than accounts of power structures (soft and hard power, political alliances, financial flows, decisions about wealth, forms of government, types of economy, ideological differences and conflicts, social connections and practices, historical antecedents such as colonialism, legal agreements, and the role of supranational corporations).

Flows of energy aren’t sufficient explanations. They are manifestations of power relations, such as new alliances (OPEC), financial might (reserve currencies), multinational corporations, availability of alternative technologies, economic mistakes, legal frameworks, and military and naval power.

Jo Di Graphics, CC BY-SA 4.0 https://creativecommons.org/licenses/by-sa/4.0, via Wikimedia Commons

Di Paolo’s combination of strict and loose approaches to autopoiesis is not a successful response to my criticism of the latter. Once strict conditions about independence from an environment are dropped, autopoiesis cannot sustain a claim to autonomy on the basis of precariousness. Any such claim will weaken explanations of processes, as I have shown in relation to power.

My turn to holistic explanation isn’t justified solely on the basis of analyses of precariousness. It also applies to Di Paolo’s dependence on the concept of closure in his definitions of autonomy.

In his work, there are no stand-alone definitions, such as ‘autonomy means X’. Instead, the definitional model is of a ladder of dependencies: ‘autonomy is understood in terms of X, understood in terms of Y, understood in terms of Z where X, Y, Z (and more) are indispensable terms of the theory’.

To demonstrate this model, I will quote a definition of autonomy in full. It is very long passage, but this is an unavoidable factor of the style of argument:

Accordingly, an autonomous system is defined as a system composed of several processes that actively generate and sustain an identity under precarious circumstances. In this context, to generate an identity is to possess the property of operational closure. This is the property that among the conditions affecting the operation of any constituent process in the system there will always be one or more processes that also belong to the system. And, in addition, every process in the system is a condition for at least one other constituent process, thus forming a network. In other words, there are no processes that are not
conditioned by other processes in the network, which does not mean, of course, that external processes cannot also influence the constituent processes, only that such processes are not part of the operationally closed network as they do not depend on the constituent processes. Similarly, there may be processes that are influenced by constituent processes but do not themselves condition any of them and are therefore not part of the operationally closed network. In their mutual dependence, the network of processes closes upon itself and defines a unity that regenerates itself (in the space where these processes occur). Precarious circumstances are those in which isolated constituent processes will tend to run down or extinguish in the absence of the organization of the system in an otherwise equivalent physical situation. In other words, individual constituent processes are not simply conditioned (e.g., modulated, adjusted, modified, or coupled to other processes), but they also depend for their continuation on the organizational network they sustain; they are enabled by it and would not be able to run isolated.

Ezekiel A. Di Paolo (2009) ‘Overcoming Autopoiesis: An Enactive Detour on the Way from Life to Society’ Autopoiesis in Organization Theory and Practice Advanced Series in Management, 43–68, p 55

Here is the ladder with partly defined concepts in bold:

  1. An autonomous system is composed of several processes
  2. These processes actively generate and sustain an identity under precarious circumstances
  3. Identity is operational closure
  4. For closure to hold one or more conditions affecting a constituent process must belong to the system
  5. Every process of the system is a condition for at least one other process of the system
  6. If closure and constituent conditions hold, the system is a network
  7. External processes can influence constituent processes but must not be conditioned by them
  8. Network processes are mutually dependent and the network therefore closes in on itself
  9. A closed network defines a unity that regenerates itself in the space defined by the processes
  10. Precarious circumstances are those where constituent processes tend to run down without the organisation of the system
  11. For example, conditioned means modulated, adjusted, modified or coupled
  12. Constituent processes are enabled bydepend upon – the organisation of the system they sustain.

The trail of definitions up the ladder is therefore: autonomy => identity => closure => condition => process => network => external influence => mutual dependence => unity => precarious circumstance => (the examples of) modulation, adjustment, modification and coupling => enabling and dependence. For the full definition of autonomy all of these terms are necessary.

Closure, closed network, influence and condition are the concepts where the critical points against Di Paolo’s claims for autonomy return. If external processes can influence the processes of a closed network, then the idea of autonomy fails. It is not enough to distinguish dependence on internal constituent processes and conditions from external influence and conditions because, as cases of control and power show, influence is a condition.

In later work, Di Paolo sometimes refers to mathematical terms for the definition of ‘network identity’ and ‘closure’, where identity is defined by set theoretical relations (if certain set theoretical relations hold through a series of operations then identity is maintained) and closure is understood as a closed field in algebra. If the roots of any polynomial having coefficients in a field K are in K, then K is closed: ‘the term “closure” is intended in the algebraic sense of a set of objects being closed under a given set of operators.’ (‘The theoretical foundations of enaction: Precariousness’, p 2)

At least for the above definition of autonomy, these mathematical terms are an imperfect match, since they would at best only be analogies, with exemplary rather than foundational roles. The analogical relation goes from descriptive terms such as process and network to highly formal ones: polynomial, function, objects, set, membership and binary operations. Nonetheless, it might be possible to give a strict mathematical translation for terms such as precarious or autonomous.

This translation is absent for the above definition. Even in the more formal work by Di Paolo and Beer, mathematical accounts are retained as models rather than sufficient formalisations and precariousness is not given a strictly mathematical formulation but rather one in terms of emergence from substrates: ‘A closure is systemically fragile to the extent that its organization is emergent from some underlying substrate.’ (‘The theoretical foundations of enaction: Precariousness’, p 4) The same is true for the use of ‘enabled’ where A is enabled by B if B is a ‘material requirement’ for A, with no direct match to a mathematical term.

The greatest weakness of the mathematical analogy is around the concept of field. For a field F to be closed there must be no polynomial with a root outside F. The field of real numbers is not closed, since the root of the equation x²+1=0 is not a real number.

Closures in the autopoietic definition of autonomy have many operations with ‘solutions/roots’ outside the closure. In short, they act on their environment beyond their boundaries or unity; that’s one of main points of the enactive theory. If an autopoietic network F has effects on its environment then, at least analogically, the closure of F is not F but rather the environment (or elements of the environment) that operations in F have effects upon.

The same problem holds for the more limited set of specific operations that define a closure. It is non-trivial to define operations for closure; for instance, the positive natural numbers are closed for addition, but are not closed for subtraction, since some subtractions give answers outside that field.

The operations of Game of Life are highly limited, those of enactive autopoiesis aren’t. Even this limitation is dubious, since from some perspectives an apparently limited real operation has wide effects, such as energy depletion (taking resources from external processes), waste products (heat, pollutants) and unexpectedly large repercussions from apparently minuscule effects (large-scale changes in behaviour from apparently innocuous innovations).

A potential answer to these objections could point to the concept of condition. However, this technical philosophical term used across many different types of argument – mainly transcendental – is only loosely defined by Di Paolo through examples of processes (modulation, adjustment, modification and coupling). For at least the first three of these, autopoietic processes are frequently conditions for their environments (in cases of clearing, building and consumption, for example).

It could be claimed that network processes don’t condition their environments in this way, since they act solely on the self-creation of elements of the network. This doesn’t pass, since those network operations still operate on the environment through intermediary processes and are successful and adopted because they do.

For instance, it could be claimed that a newly created camouflage function is strictly internal and not a condition for the environment, since it just changes the appearance of other features and processes (colour, texture, pattern). Yet those effects do modulate, adjust, modify and sometimes couple with the environment. By laying a trap, for instance.

Camouflaged Sri Lankan Leopard. By Senthiaathavan – Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=90827616

For social systems the problem is even more acute, because autopoietic transformations have frequent effects on social (and wild) environments outside those systems (a change in law; or government; or administration; or military strategy; or technical objects; or economy; or ideology; or communication). Furthermore, these effects rebound on the autopoietic process. Autonomy and closure are both lures for social processes embedded in environments.

Two final remarks. First, none of the criticisms above apply to enactive theory if autopoietic autonomy is dropped. In fact, enactivism will be much better without autopoiesis. Second, these criticisms do not apply to fictions of autonomy and independence. They may be important ways of surviving and creating oneself in a hostile environment:

Now, once again, this was something she did not doubt – she was indivisible and precious, and could be none other than herself.

She was just tired of existing and weary of all the vexations, even if they did not cause her real pain.

Marie Ndiaye, Trois femmes puissantes, Paris: Gallimard, 2009, p254 [my translation]